Purpose: Epileptiform discharges that resemble interictal spikes can be generated by slices of neocortex treated with antagonists of gamma-aminobutyric acid A (GABA(A)) receptors. These discharges can propagate horizontally for long distances. We tested the hypothesis that propagation occurs through preferred horizontal pathways that lie in a particular cortical layer. Methods: Slices were prepared from the primary somatosensory cortex of rats, maintained in vitro, and bathed with the GABA(A) receptor antagonist picrotoxin. Electrical stimuli were used to evoke single all-or-none paroxysmal field potentials (PFP) that were recorded with pairs or arrays of field potential electrodes. Results: To test which laminae are necessary for propagation, vertical cuts were made to force the PFP to spread horizontally through particular layers. If slices were bathed in a high dose of picrotoxin (35 mu M), a bridge of cortex 350 mu m thick placed at any lamina was sufficient to support PFP propagation. However, in low picrotoxin doses (2.5 mu M), similarly sized bridges had to include tissue from layers 4/5 or 5/6 to support propagation. When slices were cut horizontally (i.e., parallel to the pia) in strips, either upper-, middle-, or lower-layer strips were sufficient to support PFP propagation if the picrotoxin concentration was high; however, in low picrotoxin doses, only horizontal strips that included layer 5 could support propagation. Finally, in intact picrotoxin-treated slices, focal applications of GABA were systematically applied to different laminae as the PFP propagated past; GABA was most effective at blocking or delaying propagation when it was applied to layer 5b. Conclusions: We conclude that epileptiform propagation can occur through a variety of horizontal pathways when cortical inhibition is strongly impaired. However, when inhibition is reduced only moderately, axonal pathways in layer 5 are critical for seizure spread.